Appears in Proceedings of the SIGIR - 99 Workshop on Recommender Systems : Algorithms and Evaluation , 1999

Recommender systems improve access to relevant products and information by making personalized suggestions based on previous examples of a user's likes and dislikes. Most existing recommender systems use social ltering methods that base recommendations on other users' preferences. By contrast , content-based methods use information about an item itself to make suggestions. This approach has the advantage of being able to recommended previously unrated items to users with unique interests and to provide explanations for its recommendations. We describe a content-based book recommending system that utilizes information extraction and a machine-learning algorithm for text categorization. Initial experimental results demonstrate that this approach can produce accurate recommendations. These experiments are based on ratings from random samplings of items and we discuss problems with previous experiments that employ skewed samples of user-selected examples to evaluate performance. 1 INTRODUCTION There is a growing interest in recommender systems that suggest music, lms, books, and other products and services to users based on examples of their likes and dislikes 20, 27, 12]. A number of successful startup companies like Fireey, Net Perceptions, and LikeMinds have formed to provide recommending technology. On-line book stores like Amazon and BarnesAndNoble have popular recommendation services, and many libraries have a long history of providing reader's advisory services 2, 22]. Such services are important since readers' preferences are often complex and not readily reduced to keywords or standard subject categories , but rather best illustrated by example. Existing recommender systems almost exclusively utilize a form of computerized matchmaking called collabora-tive or social ltering. The system maintains a database of the preferences of individual users, nds other users whose known preferences correlate signiicantly with a given pa